The New Yorker has a fascinating article on what the author calls the “supernova era” in the NBA. The article nicely describes what looks like at first sight a paradox. The number of talented players able to score in any position (especially long-range shots) and more generally to realize any offensive action efficiently has never been so high in the history of the game, and yet the game has never been so centered on some individualities who are literally carrying the offense of their teams on their shoulders. The result is that superstars are even more important, and therefore, valuable than they used to be.
What is paradoxical about this? Let’s make simple supply-and-demand reasoning. Suppose agent A has a particular skill S that is demanded in the labor market. Let’s suppose that A is among the very few to offer S. If S is demanded by many, A has a bargaining power that will make them able to receive most of the marginal value created by the use of the skill in some organizational context. The higher this value and the less S is offered in the market, the more A will receive benefits from the use of S.
Now, suppose that S becomes more and more common. Mechanically, the increase in the supply of S – assuming that the demand remains roughly the same – should reduce A’s bargaining power. At the extreme, if S becomes fairly common, the marginal value that it creates will be almost fairly captured by the demand side of the market. That doesn’t mean that S has no longer any value, but rather that people like are less valuable because they no longer offer something distinctive.
What is happening in the NBA seems to contradict the prediction of this simple model. To simplify, let’s say that S corresponds to the ability of a player to create their own shot and to score efficiently. When I started to watch the NBA in the 1990s, very few players had this ability – I would say on average not more than one per team. These players were very valuable superstars. Today, virtually almost all players have this skill. By the above reasoning, this skill should no longer discriminate between players and therefore having it should be less valuable. But this is not what is happening.
If there is any paradox here, it is only apparent. There are several reasons for this. First, the analysis simplifies things too much. The value of a player is not attached to a single skill but rather to a vector of skills, and it is unclear that significantly more players have this vector of skills today than in the past. Second, the organizational context matters. This is well explained in the New Yorker article. In the case at stake, what is happening is that the multiplication of players having the relevant skills makes it more difficult for defenses to focus on the best player of the team. The more the best player is surrounded by other talented players, the less they will be tightly guarded, making them in capacity to be even more impactful than before. More formally, that means that having the skill S is not an intrinsic property of the agent, but is rather a property of the agent in their organizational context. Finally, there is the so-called superstar effect, as famously described in a paper by the economist Sherwin Rosen.[1] The superstar effect refers to a mechanism that accounts for the non-linear relationship between talent and the benefits received. If we assume that S is no longer a binary but rather a continuous variable, the superstar effect indicates that those who have the highest value for S will receive most of the benefits created, even if this value is only a little above those of others. In the NBA case, it is certainly true that the average value of S has increased in the population of players, but that does not change the fact that relatively speaking only a few are at the top.
Let’s consider now an analogy. The emergence of LLM (large language models) and text-to-image AI will contribute to making some artistic and literary “skills” far more common. Basically, everyone will now be in the capacity to generate nice illustrations or to produce well-written texts. We could then naively believe that that will reduce the benefits of those of us who already add these skills and make valuable use of them. However, the NBA case suggests that this is probably a mistaken belief.
Being a great writer or artist does not reduce to a single skill or even to the single set of skills that AI emulates. AI is good at generating a text or an image based on a description or a question but is for a moment unhelpful to find interesting topics or ideas. Moreover, the institutional context matters. Because the opportunity costs for producing texts and illustrations are decreasing thanks to AI, we should expect a large increase in the products of literary and artistic production. While the supply increases, there is no reason however to expect that the demand will. While this should apparently reduce the price that traditional producers were able to demand, this will probably not be the case because it will make even more obvious the importance of these additional skills that are not yet well emulated by AI.
There is a significant difference with the NBA case, however. In the latter, performance (or quality) is a relatively objective measure. This is indeed part of what is called the “analytic revolution” in professional sports. The mass of data available makes it possible to finely measure the performance of a player in the different dimensions of the game and their contribution to the overall performance of the team. Quality is less easy to assess in most other domains. It is more difficult for people to navigate among the mass of products offered to determine which one is the “best.” As it is well-known, information asymmetries with respect to quality may even have adverse effects leading the good quality being wiped out from the market.
In the literary and artistic markets, this problem has traditionally been solved thanks to mechanisms such as reputation, labels, critiques, and other signaling devices. The coming increase in the production of literary and artistic products will only magnify the importance of these devices. What will be valuable is not only the fact of having the set of skills that is not emulated by AI, but of being able to signal that one has it. Far from weakening the superstar effect, this will probably strengthen it. The happy few who happen to have the desired set of skills and who already are in a good position to signal this will be in a position to increase the benefits that accrue to them, even though some of them will also be captured by the information providers that help to deal with information asymmetries.
So here is the lesson of this false paradox and imperfect analogy. While in appearance AI will open access to valuable skills, it will reinforce the inherent inequality of the economic retribution scheme for talents.
[1] Sherwin Rosen, “The Economics of Superstars,” The American Economic Review 71, no. 5 (1981): 845–58.